ABSTRACT In this proposal we are directing attention to L1, the target of higher thalamocortical connections, cortical feedback and site of input to apical dendrites of pyramidal cells (Pyr). Inputs to L1 were considered to be diffuse and non-specific until we have shown that inputs to L1 of mouse V1 are clustered55. The clustering challenged the notion that rodent visual cortex is non-columnar93 and revealed a mesoscale architecture which is preserved in primate cortex55. We have recently found a similar spatial clustering of neuropil and somata of GABAergic neurons (INT). The modularity involves PV- (parvalbumin), SOM- (somatostatin) and VIP- (vasoactive intestinal peptide) expressing cells. The clustering of INT contrasts with the uniformity envisioned by the canonical circuit model25, and suggests module-specific motifs for counterbalancing excitation with inhibition. The periodicity of INT-rich and INT-poor clusters interdigitates with thalamocortical and cortical long- range connections to L1, where they synapse with INT in source-specific fashion. The clustering of INT suggests a non-uninform distribution of inhibition across V1, which differs from previous proposals143. INT clusters resemble patches of Pyr cells with different spatiotemporal senitivities55. Preliminary results suggest that neurons with high temporal acuity are preferentially localized in INT-rich interpatches, whereas neurons with high spatial acuity reside in INT-poor patches. The overlap of high temporal acuity with INT-rich modules raises several important questions, whether: inhibition in these modules is stronger, feedforward inhibition mediated by PVs is responsible for it, the more strongly inhibited Pyr cells project to specific targets, and whether PVs play a role in the diverse spatiotemporal visual preferences in patches and interpatches. INT-rich and INT-poor modules do not exist in isolation. Only INT-poor patches receive input from the lateral geniculate nucleus and feedback from the higher visual areas, LM, AL and RL, whereas input from the lateral posterior thalamus (LP) and top-down projections to INT-rich interpatches originate from dorsal stream areas136, PM and AM. Thus, INT-rich interpatches are preferentially connected to dorsal stream areas and the LP, from where they receive attention and locomotion-related inputs used for spatial navigation and detection of unexpected motion incongruent with the running speed104,109,124. Previous studies2,70 have shown that suppressive top-down signals from the stimulus surround are mediated through SOM neurons, whereas top-down signals are mediated via VIP-cell-mediated disinhibition34,147. This suggests that top-down information for object segmentation and visually guided actions may differentially involve INT-rich and INT-poor modules. To test these hypotheses we propose to determine whether: 1) the distribution of PV, SOM and VIP neurons in V1 is modular, 2) the strength of PV-, SOM- and VIP-mediated inhibition in INT-rich and INT-poor modules is different, 3) INT-rich and INT-poor modules have different inputs and outputs, and (4) the visual sensitivities of INT-rich and INT-poor modules are differentially affected by inhibition from PV-, SOM- and VIP neurons.